The forest products industry has significant optimisation opportunities similar to many other sectors. Customer orders must be satisfied; the operation must comply with rules and regulations, particularly health, safety and environment, and production costs must be controlled and minimised.
To succeed, managers and supervisors must make the right decisions in a complex, ever-changing environment. While traditional optimisation technology can create plans, it has faced challenges in its implementation. It cannot deal with real-world complexity or respond rapidly to change. New technology based on artificial intelligence (AI) overcomes these problems and has the potential to revolutionise the planning and operation of the forest product supply chain.
FROM UNIVERSITY RESEARCH TO GLOBAL SUCCESS
The Opturion story started at Imperial College London, where Professor Mark Wallace (now research director at Opturion) led a team of researchers to develop a new form of optimisation, constraint programming, part of the AI optimisation branch.
One very successful project was with CISCO for internet traffic optimisation. It was so successful that CISCO adopted the technology in 2004. At the same time, Prof Wallace moved to Monash University in Australia to continue his research. A further seven years of work paved the way to form Opturion, a university spin-out facilitated by the Commonwealth Scientific and Industrial Research Organisation.
Ten years later, Opturion has many large customers in Australia and operations in the UK and Chile. The company has successfully applied its technology in supply chains (manufacturing, transport and logistics) and, more recently, new energy.
HOW AI IS OPTIMISING SUPPLY CHAINS
Unlike traditional approaches such as mixed-integer linear programming (MILP), constraint programming does not need to approximate or simplify the business rules and relationships that define the problem. Instead, it takes a very pragmatic approach and seeks legal and valid solutions (called feasible solutions). Once it finds a feasible solution, the algorithm proceeds to find better and better solutions. After a predetermined time or other criterion, the process finishes. Consistency of time is a crucial advantage as the solution time of MILP is unreliable, if it solves at all. Another quality is the ability to quickly produce a feasible solution, essentially to re-optimise when things change. Finally, the technology is scalable to address large problems in a reasonable time. By their nature, supply chains are complex, large and dynamic, meaning that AI-based optimisation will naturally become the technology of choice.
HOW OPTIMISATION WORKS ACROSS OTHER INDUSTRIES
Supply chain optimisation is a wellestablished technology in industries such as bulk chemicals, petrochemicals, mass production and food and beverages. The predominant application is tactical planning, typically aggregating into daily or weekly ‘buckets’ where the traditional approaches work best. Opturion has taken this to the next level in two aspects:
- We are providing solutions with detailed schedules. As well as day-by-day, the optimiser can produce hour-by-hour or even minute-by-minute solutions. This ability offers opportunities for better equipment co-ordination and consideration of driver breaks, for example.
- We use the same technology to provide long-term, medium-term, and operational optimisation within the same environment, reducing development and maintenance costs and delivering consistent results.
This new capability has enabled new solutions in more complex or dynamic areas:
- Sectors with highly variable and dynamic supply chains include advanced manufacturing, make-to-order, specialty retail and last-mile logistics.
- Complex supply chains, like fuels and bulk liquids, have challenges with just-in-time, scheduling and routing, and individual vehicle load planning considerations.
THE UNIQUE CHALLENGES OF FORESTRY OPERATIONS
Forestry operations are complex and require the co-ordination of harvesting, loading and transporting equipment to deliver different species of wood to customers such as sawmills, pulp mills and plywood production plants. Like retail, there are elements of push (production rate anticipating future demand) and pull (the customer has placed an order). Similarly, there are ‘sell-by’ dates depending on the product: harvested wood starts to dry out and split, making it unsuitable for plywood and sawmills, but this does not apply to wood for pulping. Additional challenges exist around:
- Variability in the size and weight of the wood
- Harvesting and loading rates, both of which are subject to weather conditions, the terrain, the amount of daylight and the dimensions of the wood
- The reliability of equipment working under challenging conditions
- Keeping track of the location and amount of stock in the forest
- Forest roads are often unsealed, singletrack and slow to traverse, making transit times unreliable
In essence, wood, being a natural product, is variable, and production and transport occur in physically difficult conditions over a wide geographical area.
The planning and scheduling task is a multi-faceted optimisation challenge that includes the following sub-problems:
- Sequencing at load and unload points without exceeding the capacity of the loading and unloading equipment.
- Allocating and sequencing loads onto vehicles, subject to compatibility constraints, delivery windows and driver fatigue management constraints.
- Determining product flows in cases where one product may go from the same source to multiple destinations or from various sources to the same destination.
Each of these is a complex optimisation problem and, combined, they form a unique challenge requiring an integrated optimisation solution.
OPTIMISATION IN THE TIMBER TRADE: ARAUCO CHILE
The ARAUCO project essentially upgrades a first-generation optimisation application. ARAUCO is a global manufacturer of forest products, serving customers with excellence in the manufacturing and distribution of wood products. With roots in sustainable forestry and innovative product development, ARAUCO brings more than a half-century of manufacturing and supply chain excellence to the market.
The project has benefitted immensely from ARAUCO’s experience and the infrastructure put in place to support the previous application. However, the new challenge was to provide a much richer solution, with more detailed scheduling and consideration of many more critical constraints, such as driver breaks, loading times and queuing. Another aspect of the project was to improve scalability so that ARAUCO could optimise its entire operation in one go, rather than splitting it up into parts. Solving the combined problem opened up new opportunities for better co-ordination and efficiencies.
The key objectives are:
- Utilise the available drivers, equipment and vehicles as efficiently as possible
- Create detailed schedules that avoid queueing and consider driver breaks
- Create practical schedules that have a high degree of adherence
Tactical planning and scheduling is now in daily operation, and the priority is to embed the system into business as usual, incrementally improving the solution and consolidating the expected productivity benefits. Further applications in the project plan include real-time dynamic optimisation and long-term planning.
THE TIME HORIZONS FROM LONG-TERM PLANNING TO MINUTE-TO-MINUTE OPTIMISATION
Supply chain optimisation typically operates over different time horizons:
- Enterprise Resource Planning (ERP) or Sales and Operating Plan (S&OP) driven by sales forecasts
- Tactical planning over periods from weeks to months
- Detailed scheduling, day ahead to week ahead
- Rescheduling
Over time, the problem changes. Longterm planning is more about estimating the resources, in terms of material, people and assets, that would be required to meet the sales forecast. If the forecast exceeds the available resources, we have a supply problem and need to find additional resources. Otherwise, we will have a problem with demand and will look for more customers. Tactical planning is more about allocating the available resources. Scheduling is a much more difficult challenge. Detailed schedules are required to co-ordinate upstream and downstream resources and avoid conflicts and congestion. Rescheduling requires detailed status information as a starting point to go forward. It has complex logic to express what can be changed (for example, a future pickup) and what must be left alone (a delivery from a loaded truck en route).
INTEGRATION OF OPTIMISATION
We were fortunate that ARAUCO had optimisation experience and mature business processes to support it. Customers new to optimisation need to think about change management. Moving to more automated planning and scheduling requires accurate, complete data to work correctly. Humans are very good at filling in gaps, interpreting data and memorising rules and constraints. For success, it is essential to:
- Agree on the interpretation of the rules, regulations and other constraints.
- Build and maintain a fleet and driver information database and other sources of truth that change infrequently.
- Ensure that customer orders have sufficient, accurate data
- Provide the ability to track the current situation, such as order status and inventory
It is also essential that users can build their solution and compare it with the optimiser so that they develop faith in the solution.
THE BUSINESS BENEFITS
Productivity, efficiency, or cost reduction has typically motivated optimisation. Increasing asset utilisation, trucks, drivers and cranes, and reducing empty running and waiting times result from better scheduling and co-ordination. Practical schedules that reflect reality and can be adhered to are also essential.
However, the benefits of customer service and safety are equally important. Service levels and safety rules can be coded as hard constraints, never to be violated. If there is discretion, we can use soft limitations and warnings. Either way, customer service and safety are kept front and centre in the scheduling task and never forgotten.
Actual examples of benefits for more complex, dynamic operations include:
- 10% increase in productivity for the retail supply chain
- 20% reduction in customer complaints for courier delivery
- 10% increase in productivity for advanced manufacturing
We don’t expect forest products to be that different.
FUTURE PERSPECTIVES
Re-optimisation, whereby the frequency is more often than once a day, is just the start. Our experience of dispatch, where the optimiser runs every minute or so, suggests that additional benefits are possible by using additional data streams. This development has the potential for automated operation without a planner or allocator. The system directs the operation.
Data fusion will enable more accurate schedules with up-to-date information on stock levels, harvesting rates, vehicle location and queues.
The last piece of the jigsaw is machine learning, where equipment and human performance in the field, such as loading and unloading and queueing behaviour, can be analysed, modelled and predicted to enhance schedule adherence even further.
CONCLUSIONS
Forest product supply chains have unique characteristics but many things in common with those in retail and advanced manufacturing – complexity and variability present challenges and opportunities for improvements in productivity, customer service and compliance. New optimisation technology based on AI is proving to be up to this challenge and delivers accurate, detailed schedules with high adherence. The ARAUCO project is a case in point.
